~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv4/netfilter/ipt_CLUSTERIP.c

Version: ~ [ linux-6.1-rc7 ] ~ [ linux-6.0.10 ] ~ [ linux-5.19.17 ] ~ [ linux-5.18.19 ] ~ [ linux-5.17.15 ] ~ [ linux-5.16.20 ] ~ [ linux-5.15.80 ] ~ [ linux-5.14.21 ] ~ [ linux-5.13.19 ] ~ [ linux-5.12.19 ] ~ [ linux-5.11.22 ] ~ [ linux-5.10.156 ] ~ [ linux-5.9.16 ] ~ [ linux-5.8.18 ] ~ [ linux-5.7.19 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.225 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.267 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.300 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.334 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.302 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.9 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /* Cluster IP hashmark target
  2  * (C) 2003-2004 by Harald Welte <laforge@netfilter.org>
  3  * based on ideas of Fabio Olive Leite <olive@unixforge.org>
  4  *
  5  * Development of this code funded by SuSE Linux AG, http://www.suse.com/
  6  *
  7  * This program is free software; you can redistribute it and/or modify
  8  * it under the terms of the GNU General Public License version 2 as
  9  * published by the Free Software Foundation.
 10  *
 11  */
 12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 13 #include <linux/module.h>
 14 #include <linux/proc_fs.h>
 15 #include <linux/jhash.h>
 16 #include <linux/bitops.h>
 17 #include <linux/skbuff.h>
 18 #include <linux/slab.h>
 19 #include <linux/ip.h>
 20 #include <linux/tcp.h>
 21 #include <linux/udp.h>
 22 #include <linux/icmp.h>
 23 #include <linux/if_arp.h>
 24 #include <linux/seq_file.h>
 25 #include <linux/refcount.h>
 26 #include <linux/netfilter_arp.h>
 27 #include <linux/netfilter/x_tables.h>
 28 #include <linux/netfilter_ipv4/ip_tables.h>
 29 #include <linux/netfilter_ipv4/ipt_CLUSTERIP.h>
 30 #include <net/netfilter/nf_conntrack.h>
 31 #include <net/net_namespace.h>
 32 #include <net/netns/generic.h>
 33 #include <net/checksum.h>
 34 #include <net/ip.h>
 35 
 36 #define CLUSTERIP_VERSION "0.8"
 37 
 38 MODULE_LICENSE("GPL");
 39 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 40 MODULE_DESCRIPTION("Xtables: CLUSTERIP target");
 41 
 42 struct clusterip_config {
 43         struct list_head list;                  /* list of all configs */
 44         refcount_t refcount;                    /* reference count */
 45         refcount_t entries;                     /* number of entries/rules
 46                                                  * referencing us */
 47 
 48         __be32 clusterip;                       /* the IP address */
 49         u_int8_t clustermac[ETH_ALEN];          /* the MAC address */
 50         int ifindex;                            /* device ifindex */
 51         u_int16_t num_total_nodes;              /* total number of nodes */
 52         unsigned long local_nodes;              /* node number array */
 53 
 54 #ifdef CONFIG_PROC_FS
 55         struct proc_dir_entry *pde;             /* proc dir entry */
 56 #endif
 57         enum clusterip_hashmode hash_mode;      /* which hashing mode */
 58         u_int32_t hash_initval;                 /* hash initialization */
 59         struct rcu_head rcu;
 60 
 61         char ifname[IFNAMSIZ];                  /* device ifname */
 62         struct notifier_block notifier;         /* refresh c->ifindex in it */
 63 };
 64 
 65 #ifdef CONFIG_PROC_FS
 66 static const struct file_operations clusterip_proc_fops;
 67 #endif
 68 
 69 static unsigned int clusterip_net_id __read_mostly;
 70 
 71 struct clusterip_net {
 72         struct list_head configs;
 73         /* lock protects the configs list */
 74         spinlock_t lock;
 75 
 76 #ifdef CONFIG_PROC_FS
 77         struct proc_dir_entry *procdir;
 78 #endif
 79 };
 80 
 81 static inline void
 82 clusterip_config_get(struct clusterip_config *c)
 83 {
 84         refcount_inc(&c->refcount);
 85 }
 86 
 87 
 88 static void clusterip_config_rcu_free(struct rcu_head *head)
 89 {
 90         kfree(container_of(head, struct clusterip_config, rcu));
 91 }
 92 
 93 static inline void
 94 clusterip_config_put(struct clusterip_config *c)
 95 {
 96         if (refcount_dec_and_test(&c->refcount))
 97                 call_rcu_bh(&c->rcu, clusterip_config_rcu_free);
 98 }
 99 
100 /* decrease the count of entries using/referencing this config.  If last
101  * entry(rule) is removed, remove the config from lists, but don't free it
102  * yet, since proc-files could still be holding references */
103 static inline void
104 clusterip_config_entry_put(struct net *net, struct clusterip_config *c)
105 {
106         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
107 
108         local_bh_disable();
109         if (refcount_dec_and_lock(&c->entries, &cn->lock)) {
110                 /* In case anyone still accesses the file, the open/close
111                  * functions are also incrementing the refcount on their own,
112                  * so it's safe to remove the entry even if it's in use. */
113 #ifdef CONFIG_PROC_FS
114                 if (cn->procdir)
115                         proc_remove(c->pde);
116 #endif
117                 list_del_rcu(&c->list);
118                 spin_unlock(&cn->lock);
119                 local_bh_enable();
120 
121                 unregister_netdevice_notifier(&c->notifier);
122 
123                 return;
124         }
125         local_bh_enable();
126 }
127 
128 static struct clusterip_config *
129 __clusterip_config_find(struct net *net, __be32 clusterip)
130 {
131         struct clusterip_config *c;
132         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
133 
134         list_for_each_entry_rcu(c, &cn->configs, list) {
135                 if (c->clusterip == clusterip)
136                         return c;
137         }
138 
139         return NULL;
140 }
141 
142 static inline struct clusterip_config *
143 clusterip_config_find_get(struct net *net, __be32 clusterip, int entry)
144 {
145         struct clusterip_config *c;
146 
147         rcu_read_lock_bh();
148         c = __clusterip_config_find(net, clusterip);
149         if (c) {
150 #ifdef CONFIG_PROC_FS
151                 if (!c->pde)
152                         c = NULL;
153                 else
154 #endif
155                 if (unlikely(!refcount_inc_not_zero(&c->refcount)))
156                         c = NULL;
157                 else if (entry)
158                         refcount_inc(&c->entries);
159         }
160         rcu_read_unlock_bh();
161 
162         return c;
163 }
164 
165 static void
166 clusterip_config_init_nodelist(struct clusterip_config *c,
167                                const struct ipt_clusterip_tgt_info *i)
168 {
169         int n;
170 
171         for (n = 0; n < i->num_local_nodes; n++)
172                 set_bit(i->local_nodes[n] - 1, &c->local_nodes);
173 }
174 
175 static int
176 clusterip_netdev_event(struct notifier_block *this, unsigned long event,
177                        void *ptr)
178 {
179         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
180         struct clusterip_config *c;
181 
182         c = container_of(this, struct clusterip_config, notifier);
183         switch (event) {
184         case NETDEV_REGISTER:
185                 if (!strcmp(dev->name, c->ifname)) {
186                         c->ifindex = dev->ifindex;
187                         dev_mc_add(dev, c->clustermac);
188                 }
189                 break;
190         case NETDEV_UNREGISTER:
191                 if (dev->ifindex == c->ifindex) {
192                         dev_mc_del(dev, c->clustermac);
193                         c->ifindex = -1;
194                 }
195                 break;
196         case NETDEV_CHANGENAME:
197                 if (!strcmp(dev->name, c->ifname)) {
198                         c->ifindex = dev->ifindex;
199                         dev_mc_add(dev, c->clustermac);
200                 } else if (dev->ifindex == c->ifindex) {
201                         dev_mc_del(dev, c->clustermac);
202                         c->ifindex = -1;
203                 }
204                 break;
205         }
206 
207         return NOTIFY_DONE;
208 }
209 
210 static struct clusterip_config *
211 clusterip_config_init(struct net *net, const struct ipt_clusterip_tgt_info *i,
212                       __be32 ip, const char *iniface)
213 {
214         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
215         struct clusterip_config *c;
216         int err;
217 
218         c = kzalloc(sizeof(*c), GFP_ATOMIC);
219         if (!c)
220                 return ERR_PTR(-ENOMEM);
221 
222         strcpy(c->ifname, iniface);
223         c->ifindex = -1;
224         c->clusterip = ip;
225         memcpy(&c->clustermac, &i->clustermac, ETH_ALEN);
226         c->num_total_nodes = i->num_total_nodes;
227         clusterip_config_init_nodelist(c, i);
228         c->hash_mode = i->hash_mode;
229         c->hash_initval = i->hash_initval;
230         refcount_set(&c->refcount, 1);
231         refcount_set(&c->entries, 1);
232 
233         spin_lock_bh(&cn->lock);
234         if (__clusterip_config_find(net, ip)) {
235                 spin_unlock_bh(&cn->lock);
236                 kfree(c);
237 
238                 return ERR_PTR(-EBUSY);
239         }
240 
241         list_add_rcu(&c->list, &cn->configs);
242         spin_unlock_bh(&cn->lock);
243 
244 #ifdef CONFIG_PROC_FS
245         {
246                 char buffer[16];
247 
248                 /* create proc dir entry */
249                 sprintf(buffer, "%pI4", &ip);
250                 c->pde = proc_create_data(buffer, S_IWUSR|S_IRUSR,
251                                           cn->procdir,
252                                           &clusterip_proc_fops, c);
253                 if (!c->pde) {
254                         err = -ENOMEM;
255                         goto err;
256                 }
257         }
258 #endif
259 
260         c->notifier.notifier_call = clusterip_netdev_event;
261         err = register_netdevice_notifier(&c->notifier);
262         if (!err)
263                 return c;
264 
265 #ifdef CONFIG_PROC_FS
266         proc_remove(c->pde);
267 err:
268 #endif
269         spin_lock_bh(&cn->lock);
270         list_del_rcu(&c->list);
271         spin_unlock_bh(&cn->lock);
272         kfree(c);
273 
274         return ERR_PTR(err);
275 }
276 
277 #ifdef CONFIG_PROC_FS
278 static int
279 clusterip_add_node(struct clusterip_config *c, u_int16_t nodenum)
280 {
281 
282         if (nodenum == 0 ||
283             nodenum > c->num_total_nodes)
284                 return 1;
285 
286         /* check if we already have this number in our bitfield */
287         if (test_and_set_bit(nodenum - 1, &c->local_nodes))
288                 return 1;
289 
290         return 0;
291 }
292 
293 static bool
294 clusterip_del_node(struct clusterip_config *c, u_int16_t nodenum)
295 {
296         if (nodenum == 0 ||
297             nodenum > c->num_total_nodes)
298                 return true;
299 
300         if (test_and_clear_bit(nodenum - 1, &c->local_nodes))
301                 return false;
302 
303         return true;
304 }
305 #endif
306 
307 static inline u_int32_t
308 clusterip_hashfn(const struct sk_buff *skb,
309                  const struct clusterip_config *config)
310 {
311         const struct iphdr *iph = ip_hdr(skb);
312         unsigned long hashval;
313         u_int16_t sport = 0, dport = 0;
314         int poff;
315 
316         poff = proto_ports_offset(iph->protocol);
317         if (poff >= 0) {
318                 const u_int16_t *ports;
319                 u16 _ports[2];
320 
321                 ports = skb_header_pointer(skb, iph->ihl * 4 + poff, 4, _ports);
322                 if (ports) {
323                         sport = ports[0];
324                         dport = ports[1];
325                 }
326         } else {
327                 net_info_ratelimited("unknown protocol %u\n", iph->protocol);
328         }
329 
330         switch (config->hash_mode) {
331         case CLUSTERIP_HASHMODE_SIP:
332                 hashval = jhash_1word(ntohl(iph->saddr),
333                                       config->hash_initval);
334                 break;
335         case CLUSTERIP_HASHMODE_SIP_SPT:
336                 hashval = jhash_2words(ntohl(iph->saddr), sport,
337                                        config->hash_initval);
338                 break;
339         case CLUSTERIP_HASHMODE_SIP_SPT_DPT:
340                 hashval = jhash_3words(ntohl(iph->saddr), sport, dport,
341                                        config->hash_initval);
342                 break;
343         default:
344                 /* to make gcc happy */
345                 hashval = 0;
346                 /* This cannot happen, unless the check function wasn't called
347                  * at rule load time */
348                 pr_info("unknown mode %u\n", config->hash_mode);
349                 BUG();
350                 break;
351         }
352 
353         /* node numbers are 1..n, not 0..n */
354         return reciprocal_scale(hashval, config->num_total_nodes) + 1;
355 }
356 
357 static inline int
358 clusterip_responsible(const struct clusterip_config *config, u_int32_t hash)
359 {
360         return test_bit(hash - 1, &config->local_nodes);
361 }
362 
363 /***********************************************************************
364  * IPTABLES TARGET
365  ***********************************************************************/
366 
367 static unsigned int
368 clusterip_tg(struct sk_buff *skb, const struct xt_action_param *par)
369 {
370         const struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
371         struct nf_conn *ct;
372         enum ip_conntrack_info ctinfo;
373         u_int32_t hash;
374 
375         /* don't need to clusterip_config_get() here, since refcount
376          * is only decremented by destroy() - and ip_tables guarantees
377          * that the ->target() function isn't called after ->destroy() */
378 
379         ct = nf_ct_get(skb, &ctinfo);
380         if (ct == NULL)
381                 return NF_DROP;
382 
383         /* special case: ICMP error handling. conntrack distinguishes between
384          * error messages (RELATED) and information requests (see below) */
385         if (ip_hdr(skb)->protocol == IPPROTO_ICMP &&
386             (ctinfo == IP_CT_RELATED ||
387              ctinfo == IP_CT_RELATED_REPLY))
388                 return XT_CONTINUE;
389 
390         /* ip_conntrack_icmp guarantees us that we only have ICMP_ECHO,
391          * TIMESTAMP, INFO_REQUEST or ADDRESS type icmp packets from here
392          * on, which all have an ID field [relevant for hashing]. */
393 
394         hash = clusterip_hashfn(skb, cipinfo->config);
395 
396         switch (ctinfo) {
397         case IP_CT_NEW:
398                 ct->mark = hash;
399                 break;
400         case IP_CT_RELATED:
401         case IP_CT_RELATED_REPLY:
402                 /* FIXME: we don't handle expectations at the moment.
403                  * They can arrive on a different node than
404                  * the master connection (e.g. FTP passive mode) */
405         case IP_CT_ESTABLISHED:
406         case IP_CT_ESTABLISHED_REPLY:
407                 break;
408         default:                        /* Prevent gcc warnings */
409                 break;
410         }
411 
412 #ifdef DEBUG
413         nf_ct_dump_tuple_ip(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
414 #endif
415         pr_debug("hash=%u ct_hash=%u ", hash, ct->mark);
416         if (!clusterip_responsible(cipinfo->config, hash)) {
417                 pr_debug("not responsible\n");
418                 return NF_DROP;
419         }
420         pr_debug("responsible\n");
421 
422         /* despite being received via linklayer multicast, this is
423          * actually a unicast IP packet. TCP doesn't like PACKET_MULTICAST */
424         skb->pkt_type = PACKET_HOST;
425 
426         return XT_CONTINUE;
427 }
428 
429 static int clusterip_tg_check(const struct xt_tgchk_param *par)
430 {
431         struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
432         const struct ipt_entry *e = par->entryinfo;
433         struct clusterip_config *config;
434         int ret, i;
435 
436         if (par->nft_compat) {
437                 pr_err("cannot use CLUSTERIP target from nftables compat\n");
438                 return -EOPNOTSUPP;
439         }
440 
441         if (cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP &&
442             cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP_SPT &&
443             cipinfo->hash_mode != CLUSTERIP_HASHMODE_SIP_SPT_DPT) {
444                 pr_info("unknown mode %u\n", cipinfo->hash_mode);
445                 return -EINVAL;
446 
447         }
448         if (e->ip.dmsk.s_addr != htonl(0xffffffff) ||
449             e->ip.dst.s_addr == 0) {
450                 pr_info("Please specify destination IP\n");
451                 return -EINVAL;
452         }
453         if (cipinfo->num_local_nodes > ARRAY_SIZE(cipinfo->local_nodes)) {
454                 pr_info("bad num_local_nodes %u\n", cipinfo->num_local_nodes);
455                 return -EINVAL;
456         }
457         for (i = 0; i < cipinfo->num_local_nodes; i++) {
458                 if (cipinfo->local_nodes[i] - 1 >=
459                     sizeof(config->local_nodes) * 8) {
460                         pr_info("bad local_nodes[%d] %u\n",
461                                 i, cipinfo->local_nodes[i]);
462                         return -EINVAL;
463                 }
464         }
465 
466         config = clusterip_config_find_get(par->net, e->ip.dst.s_addr, 1);
467         if (!config) {
468                 if (!(cipinfo->flags & CLUSTERIP_FLAG_NEW)) {
469                         pr_info("no config found for %pI4, need 'new'\n",
470                                 &e->ip.dst.s_addr);
471                         return -EINVAL;
472                 } else {
473                         struct net_device *dev;
474 
475                         if (e->ip.iniface[0] == '\0') {
476                                 pr_info("Please specify an interface name\n");
477                                 return -EINVAL;
478                         }
479 
480                         dev = dev_get_by_name(par->net, e->ip.iniface);
481                         if (!dev) {
482                                 pr_info("no such interface %s\n",
483                                         e->ip.iniface);
484                                 return -ENOENT;
485                         }
486                         dev_put(dev);
487 
488                         config = clusterip_config_init(par->net, cipinfo,
489                                                        e->ip.dst.s_addr,
490                                                        e->ip.iniface);
491                         if (IS_ERR(config))
492                                 return PTR_ERR(config);
493                 }
494         }
495         cipinfo->config = config;
496 
497         ret = nf_ct_netns_get(par->net, par->family);
498         if (ret < 0)
499                 pr_info("cannot load conntrack support for proto=%u\n",
500                         par->family);
501 
502         if (!par->net->xt.clusterip_deprecated_warning) {
503                 pr_info("ipt_CLUSTERIP is deprecated and it will removed soon, "
504                         "use xt_cluster instead\n");
505                 par->net->xt.clusterip_deprecated_warning = true;
506         }
507 
508         return ret;
509 }
510 
511 /* drop reference count of cluster config when rule is deleted */
512 static void clusterip_tg_destroy(const struct xt_tgdtor_param *par)
513 {
514         const struct ipt_clusterip_tgt_info *cipinfo = par->targinfo;
515 
516         /* if no more entries are referencing the config, remove it
517          * from the list and destroy the proc entry */
518         clusterip_config_entry_put(par->net, cipinfo->config);
519 
520         clusterip_config_put(cipinfo->config);
521 
522         nf_ct_netns_put(par->net, par->family);
523 }
524 
525 #ifdef CONFIG_COMPAT
526 struct compat_ipt_clusterip_tgt_info
527 {
528         u_int32_t       flags;
529         u_int8_t        clustermac[6];
530         u_int16_t       num_total_nodes;
531         u_int16_t       num_local_nodes;
532         u_int16_t       local_nodes[CLUSTERIP_MAX_NODES];
533         u_int32_t       hash_mode;
534         u_int32_t       hash_initval;
535         compat_uptr_t   config;
536 };
537 #endif /* CONFIG_COMPAT */
538 
539 static struct xt_target clusterip_tg_reg __read_mostly = {
540         .name           = "CLUSTERIP",
541         .family         = NFPROTO_IPV4,
542         .target         = clusterip_tg,
543         .checkentry     = clusterip_tg_check,
544         .destroy        = clusterip_tg_destroy,
545         .targetsize     = sizeof(struct ipt_clusterip_tgt_info),
546         .usersize       = offsetof(struct ipt_clusterip_tgt_info, config),
547 #ifdef CONFIG_COMPAT
548         .compatsize     = sizeof(struct compat_ipt_clusterip_tgt_info),
549 #endif /* CONFIG_COMPAT */
550         .me             = THIS_MODULE
551 };
552 
553 
554 /***********************************************************************
555  * ARP MANGLING CODE
556  ***********************************************************************/
557 
558 /* hardcoded for 48bit ethernet and 32bit ipv4 addresses */
559 struct arp_payload {
560         u_int8_t src_hw[ETH_ALEN];
561         __be32 src_ip;
562         u_int8_t dst_hw[ETH_ALEN];
563         __be32 dst_ip;
564 } __packed;
565 
566 #ifdef DEBUG
567 static void arp_print(struct arp_payload *payload)
568 {
569 #define HBUFFERLEN 30
570         char hbuffer[HBUFFERLEN];
571         int j, k;
572 
573         for (k = 0, j = 0; k < HBUFFERLEN - 3 && j < ETH_ALEN; j++) {
574                 hbuffer[k++] = hex_asc_hi(payload->src_hw[j]);
575                 hbuffer[k++] = hex_asc_lo(payload->src_hw[j]);
576                 hbuffer[k++] = ':';
577         }
578         hbuffer[--k] = '\0';
579 
580         pr_debug("src %pI4@%s, dst %pI4\n",
581                  &payload->src_ip, hbuffer, &payload->dst_ip);
582 }
583 #endif
584 
585 static unsigned int
586 arp_mangle(void *priv,
587            struct sk_buff *skb,
588            const struct nf_hook_state *state)
589 {
590         struct arphdr *arp = arp_hdr(skb);
591         struct arp_payload *payload;
592         struct clusterip_config *c;
593         struct net *net = state->net;
594 
595         /* we don't care about non-ethernet and non-ipv4 ARP */
596         if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
597             arp->ar_pro != htons(ETH_P_IP) ||
598             arp->ar_pln != 4 || arp->ar_hln != ETH_ALEN)
599                 return NF_ACCEPT;
600 
601         /* we only want to mangle arp requests and replies */
602         if (arp->ar_op != htons(ARPOP_REPLY) &&
603             arp->ar_op != htons(ARPOP_REQUEST))
604                 return NF_ACCEPT;
605 
606         payload = (void *)(arp+1);
607 
608         /* if there is no clusterip configuration for the arp reply's
609          * source ip, we don't want to mangle it */
610         c = clusterip_config_find_get(net, payload->src_ip, 0);
611         if (!c)
612                 return NF_ACCEPT;
613 
614         /* normally the linux kernel always replies to arp queries of
615          * addresses on different interfacs.  However, in the CLUSTERIP case
616          * this wouldn't work, since we didn't subscribe the mcast group on
617          * other interfaces */
618         if (c->ifindex != state->out->ifindex) {
619                 pr_debug("not mangling arp reply on different interface: cip'%d'-skb'%d'\n",
620                          c->ifindex, state->out->ifindex);
621                 clusterip_config_put(c);
622                 return NF_ACCEPT;
623         }
624 
625         /* mangle reply hardware address */
626         memcpy(payload->src_hw, c->clustermac, arp->ar_hln);
627 
628 #ifdef DEBUG
629         pr_debug("mangled arp reply: ");
630         arp_print(payload);
631 #endif
632 
633         clusterip_config_put(c);
634 
635         return NF_ACCEPT;
636 }
637 
638 static const struct nf_hook_ops cip_arp_ops = {
639         .hook = arp_mangle,
640         .pf = NFPROTO_ARP,
641         .hooknum = NF_ARP_OUT,
642         .priority = -1
643 };
644 
645 /***********************************************************************
646  * PROC DIR HANDLING
647  ***********************************************************************/
648 
649 #ifdef CONFIG_PROC_FS
650 
651 struct clusterip_seq_position {
652         unsigned int pos;       /* position */
653         unsigned int weight;    /* number of bits set == size */
654         unsigned int bit;       /* current bit */
655         unsigned long val;      /* current value */
656 };
657 
658 static void *clusterip_seq_start(struct seq_file *s, loff_t *pos)
659 {
660         struct clusterip_config *c = s->private;
661         unsigned int weight;
662         u_int32_t local_nodes;
663         struct clusterip_seq_position *idx;
664 
665         /* FIXME: possible race */
666         local_nodes = c->local_nodes;
667         weight = hweight32(local_nodes);
668         if (*pos >= weight)
669                 return NULL;
670 
671         idx = kmalloc(sizeof(struct clusterip_seq_position), GFP_KERNEL);
672         if (!idx)
673                 return ERR_PTR(-ENOMEM);
674 
675         idx->pos = *pos;
676         idx->weight = weight;
677         idx->bit = ffs(local_nodes);
678         idx->val = local_nodes;
679         clear_bit(idx->bit - 1, &idx->val);
680 
681         return idx;
682 }
683 
684 static void *clusterip_seq_next(struct seq_file *s, void *v, loff_t *pos)
685 {
686         struct clusterip_seq_position *idx = v;
687 
688         *pos = ++idx->pos;
689         if (*pos >= idx->weight) {
690                 kfree(v);
691                 return NULL;
692         }
693         idx->bit = ffs(idx->val);
694         clear_bit(idx->bit - 1, &idx->val);
695         return idx;
696 }
697 
698 static void clusterip_seq_stop(struct seq_file *s, void *v)
699 {
700         if (!IS_ERR(v))
701                 kfree(v);
702 }
703 
704 static int clusterip_seq_show(struct seq_file *s, void *v)
705 {
706         struct clusterip_seq_position *idx = v;
707 
708         if (idx->pos != 0)
709                 seq_putc(s, ',');
710 
711         seq_printf(s, "%u", idx->bit);
712 
713         if (idx->pos == idx->weight - 1)
714                 seq_putc(s, '\n');
715 
716         return 0;
717 }
718 
719 static const struct seq_operations clusterip_seq_ops = {
720         .start  = clusterip_seq_start,
721         .next   = clusterip_seq_next,
722         .stop   = clusterip_seq_stop,
723         .show   = clusterip_seq_show,
724 };
725 
726 static int clusterip_proc_open(struct inode *inode, struct file *file)
727 {
728         int ret = seq_open(file, &clusterip_seq_ops);
729 
730         if (!ret) {
731                 struct seq_file *sf = file->private_data;
732                 struct clusterip_config *c = PDE_DATA(inode);
733 
734                 sf->private = c;
735 
736                 clusterip_config_get(c);
737         }
738 
739         return ret;
740 }
741 
742 static int clusterip_proc_release(struct inode *inode, struct file *file)
743 {
744         struct clusterip_config *c = PDE_DATA(inode);
745         int ret;
746 
747         ret = seq_release(inode, file);
748 
749         if (!ret)
750                 clusterip_config_put(c);
751 
752         return ret;
753 }
754 
755 static ssize_t clusterip_proc_write(struct file *file, const char __user *input,
756                                 size_t size, loff_t *ofs)
757 {
758         struct clusterip_config *c = PDE_DATA(file_inode(file));
759 #define PROC_WRITELEN   10
760         char buffer[PROC_WRITELEN+1];
761         unsigned long nodenum;
762         int rc;
763 
764         if (size > PROC_WRITELEN)
765                 return -EIO;
766         if (copy_from_user(buffer, input, size))
767                 return -EFAULT;
768         buffer[size] = 0;
769 
770         if (*buffer == '+') {
771                 rc = kstrtoul(buffer+1, 10, &nodenum);
772                 if (rc)
773                         return rc;
774                 if (clusterip_add_node(c, nodenum))
775                         return -ENOMEM;
776         } else if (*buffer == '-') {
777                 rc = kstrtoul(buffer+1, 10, &nodenum);
778                 if (rc)
779                         return rc;
780                 if (clusterip_del_node(c, nodenum))
781                         return -ENOENT;
782         } else
783                 return -EIO;
784 
785         return size;
786 }
787 
788 static const struct file_operations clusterip_proc_fops = {
789         .owner   = THIS_MODULE,
790         .open    = clusterip_proc_open,
791         .read    = seq_read,
792         .write   = clusterip_proc_write,
793         .llseek  = seq_lseek,
794         .release = clusterip_proc_release,
795 };
796 
797 #endif /* CONFIG_PROC_FS */
798 
799 static int clusterip_net_init(struct net *net)
800 {
801         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
802         int ret;
803 
804         INIT_LIST_HEAD(&cn->configs);
805 
806         spin_lock_init(&cn->lock);
807 
808         ret = nf_register_net_hook(net, &cip_arp_ops);
809         if (ret < 0)
810                 return ret;
811 
812 #ifdef CONFIG_PROC_FS
813         cn->procdir = proc_mkdir("ipt_CLUSTERIP", net->proc_net);
814         if (!cn->procdir) {
815                 nf_unregister_net_hook(net, &cip_arp_ops);
816                 pr_err("Unable to proc dir entry\n");
817                 return -ENOMEM;
818         }
819 #endif /* CONFIG_PROC_FS */
820 
821         return 0;
822 }
823 
824 static void clusterip_net_exit(struct net *net)
825 {
826         struct clusterip_net *cn = net_generic(net, clusterip_net_id);
827 #ifdef CONFIG_PROC_FS
828         proc_remove(cn->procdir);
829         cn->procdir = NULL;
830 #endif
831         nf_unregister_net_hook(net, &cip_arp_ops);
832         WARN_ON_ONCE(!list_empty(&cn->configs));
833 }
834 
835 static struct pernet_operations clusterip_net_ops = {
836         .init = clusterip_net_init,
837         .exit = clusterip_net_exit,
838         .id   = &clusterip_net_id,
839         .size = sizeof(struct clusterip_net),
840 };
841 
842 static int __init clusterip_tg_init(void)
843 {
844         int ret;
845 
846         ret = register_pernet_subsys(&clusterip_net_ops);
847         if (ret < 0)
848                 return ret;
849 
850         ret = xt_register_target(&clusterip_tg_reg);
851         if (ret < 0)
852                 goto cleanup_subsys;
853 
854         pr_info("ClusterIP Version %s loaded successfully\n",
855                 CLUSTERIP_VERSION);
856 
857         return 0;
858 
859 cleanup_subsys:
860         unregister_pernet_subsys(&clusterip_net_ops);
861         return ret;
862 }
863 
864 static void __exit clusterip_tg_exit(void)
865 {
866         pr_info("ClusterIP Version %s unloading\n", CLUSTERIP_VERSION);
867 
868         xt_unregister_target(&clusterip_tg_reg);
869         unregister_pernet_subsys(&clusterip_net_ops);
870 
871         /* Wait for completion of call_rcu_bh()'s (clusterip_config_rcu_free) */
872         rcu_barrier_bh();
873 }
874 
875 module_init(clusterip_tg_init);
876 module_exit(clusterip_tg_exit);
877 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp